Beruflich Dokumente
Kultur Dokumente
103
-.
0
P
105
106
*Although the interface Row is the sum of Rows in lines, a negative sign is used here because the
convention is positive for Row in the direction 2 to 3.
I07
r=l i = l
'3Evidently, diagonal entries in cells C24, D25, and so on, are zero.
4.8
INTER-REGIONAL DISPATCH
109
subject to
where D, is the demand of region r and I,, E, indicate the import and export into
regions r,
Q:g(min) I
Qrg IQrg(max),
i = 1,. . . ,n,,
r = 1,. . . , R ,
in which (max) and (min) indicate the maximum and minimum output capabilities
of generators, and the regional transfer limits
Irj
I
Imax
r = 1, .. . , R ,
rj,
where Zmax ,j and Emax ,j are the maximum permitted imports to and exports
from regions r = 1, R, and the limits
Q;gIQTo,
r = l , ...,R ,
i=l,
...,n,,
j=1
R, j#r
E , = ~ E , ~ ,r = l , ...,R ,
j=1
Solution: The solution obtained by the Solver in Figure 4.13 shows the minimum
cost of dispatch to be $6894.18. Only some constraints are binding. For example,
the upper limit of the third generator in subregion B and the second generator in
subregion E are binding; also, the lower limit of the first generator in subregion A
is binding.
The reader will gain further knowledge from an exercise of changing the limits
for generation and transmission flows, and by examining the associated sensitivity reports. In the following, the sensitivity report associated with Figure 4.13
illustrates the type of information that can be obtained therefrom. In fact, such information is used to price energy in subregions. Results of sensitivity are essential for
the settlement of markets of the deregulated electrical system.
110
4.8.1
Sensitivity Report
Figure 4.14 shows the sensitivity report. This report refers to cells in the study of
Figure 4.13. The reader should examine and ponder the reduced costs associated
with choice variables and shadow prices of constraints.
The most expensive generator feeding energy has a cost of $13.50 in region A,
despite some cheaper generators not being at their full outputs in other regions.
This is because the lower limit of output from that generator is a binding constraint.
In terms of shadow prices, the constraint regarding export from subregions C
and E are binding. Consequently, cheaper generators are backed off in C and E.
Therefore, the cost of supplying an additional unit of energy in subregion E is $8
(since the output of the $8.00 machine is only 30.105, well below its maximum
capability). This is shown in cell J19. Similarly, the cost of supplying the next
increment of energy in subregion C is $10. This is shown in cell J13. Continuing
the same logic, it can be confirmed that the shadow prices associated with cells 57
to J19 are the costs of supplying the next unit of energy in the subregions. Hence,
they represent the LMP of the subregions. The shadow price associated with cells
N33 and N36 indicate the penalty associated with the transfer limit constraint. For
example, the shadow price of $3.002 associated with cell N33 indicates that if the
constraint value for flow across this interface is increased by one (to -lOl), the
objective will decrease by $3.0024.
The penalties associated with shadow prices have to be examined for several
hours of system dispatch in order to ascertain the benefit that would accrue with
any transfer capacity upgrades. The cost of such upgrades, of course, has to be
juxtaposed with the benefits.
4.9.1
Given three warehouses 1, 2, and 3, and three destinations 4, 5, and 6, for the
given costs of transportation between them, find the optimal transportation policy
from warehouse to destination. The quantities available at warehouses, as well as
quantities needed at destinations, are given as indicated in the spreadsheet TransTranspor- portation.Xls.
tationA printout of the spreadsheet and its solution is shown in Figure 4.15. The quanXls
tities needed in destinations are shown in row 17 as 10, 20, and 15. The quantity
available at the three warehouses is indicated in column I as 5, 17, and 23. The costs
of transportation from each warehouse to destinations are indicated in rows 5,9, and
13. Choice variables are the quantities transported from the warehouses to destinations shown in rows 4,8, and 12. The objective of transportation cost is in cell F16.
4
4
4
112
where t i j is the cost of transportation from i to j , and the quantities moved from
n warehouses to j destinations should be such that
n
Ccij= e j ,
j = 1 ,... , m ,
i=l
where
Qj
j = 1,... , m ,
C c i j = Qi,
j=1
i=l
j=1
and
cij
2 0.
4.9
113
Solution: The final solution using the Solver is shown in the figure under discussion.
4.9.2 Integrating Natural Gas Transportation
and Electricity Generation
We now extend the above problem to include a case of fuel supply (natural gas)
and electricity generation. Even this extension is somewhat simplistic. In real
world situations, optimal fuel procurement by electricity generating companies is
a complex task. One enters into longer-term contracts with take or pay clause,
and one also deals in Options, Futures, and Spot marketsI4 for fuel procurement.
Furthermore, a diversification in sources of supply as well as in types of fuel to
avoid risks of shortage is common practice. Understandably, it is not possible to
simulate all the complications of fuel procurement in an example of a textbook.
Nevertheless, our intent here is to demonstrate how different types of optimization
problems can be integrated to simulate practical situations.
Toward that end, in this example consider three sources of natural gas connected
by a pipeline transportation network to three generating units owned by a Generator.
The spreadsheet Gas-Electric-Transportation.Xls is shown in Figure 4.16. Let Gas the heat content of the gas from the three wellheads be the same. However, the heat
transporrates of generators are different; GI generates 1 MWh per MBTU, G2 produces tation.xls
0.95 MWh per MBTU, and G3 produces 1.05 MWh per MBTU. The data shown
in row 3 incorporate these heat rates.
Discussion of Problem: The cost of transporting gas from the three wellheads to
the three generators are in rows 7, 15, and 23 in units of $/MBTU. The quantity that
can be delivered per hour (MBTU) is dependent on pipeline capacity as indicated in
rows 6, 14 and 22. The total amount of gas (in MBTU) that can be delivered from
each wellhead (per hour) is shown in column I. This represents pipeline capacity.
In terms of pollution arising from electricity generation, each generating unit
has different types of burners. Further there are differences in the composition of
natural gas from the three wellheads. Consequently, different rates of pollution
(thousand lbs per MBTU) are indicated for the generators in rows 10, 19, and 27,
respectively.
In this simple problem, we assume that there is no demand charge for gas delivery,
and that the cost of gas per MBTU, apart from the transportation charges, is the same
from three wellheads. The problem now is to determine the optimum deliveries in
I4Take or pay clauses are written into the fuel delivery contract, obligating the buyer to buy a certain
minimum quantity. Failure to buy the minimum quantity would still entail a certain payment or penalty.
Options and futures are more complicated. The buyer in a secondary market, or in a commodity
exchange, buys a paper contract assuring the delivery of a commodity at some future date at a certain fixed price. The commodity is seldom delivered, but the papers are traded and used as hedging
instruments to insulate the holder from price fluctuations.
The field of hedging is vast, and organizations employ professionals to hedge risks. An excellent source
of reference on these matters is Hull (1999).
114
Figure 4.16. Optimization of natural gas delivery through transportation network to electric
generators.
order to a) minimize the transportation charge, b) minimize the pollution given that
the owner of generating units is obligated to deliver 450 M W into the grid.
f i i , f2i, f3i
for i =
I15
subject to
+ fj2 + f j 3 = 450,
C
3
j=l
0.95
1.05
i=l
3
i=l
and
fji
2 0,
j = 1,2,3;
i = 1,2,3,
where fji is the gas flow from wellhead j to generator i, and tji is the transportation
cost associated with this flow.
For the case of minimizing pollution, (4.1) has to be replaced with a corresponding objective taking into account the pollution rates at generators for different fuels
as depicted in Figure 4.16.
Solution: The solution to the problem of minimizing transportation cost using the
Excel Solver is shown in Figure 4.16 to be $1233.81. The resultant pollution with
this strategy of gas purchase is 292.07 x lo3 lb. Similarly, the reader will find that
the optimum solution to minimize pollution (choose cell H33 to be minimized)
results in a total pollution of 242.5714 x lo3 lb. Correspondingly, the cost of gas
transportation increases. The cost of gas transportation is a degenerate solution as
explained in Section 3.4. This cost can vary from a cost of $2984.28 to $3018.92.
The reader is expected to examine sensitivity sheets resulting from these programs and ponder the reasons for the resulting shadow prices and reduced costs.
4.10 SOLUTION OF PRIMAL-DUAL PROBLEMS
In the example problems of Section 3.10.1, for simplicity, only less than or equal
type of inequalities were considered in the primal problem. However, Table 3.1
outlined rules to formulate the dual problem for all types of inequalities and equality
constraints. In the following examples, we formulate and solve the dual when the
primal problem includes greater than and less than type of inequalities, as well as
equality constraints. We study sensitivity reports of these two solutions and make
observations that correspond to the duality theorems of Section 3.10.2.
116
+ 100x2 + 200x3,
(4.2)
subject to
(4.3)
(4.4)
(4.5)
(4.6)
and
and
x3,
unspecijied.
j-~
ly4
(4.7)
The relationships shown in Table 3.1 stipulate constraints for the dual problem
as follows. In the primal, since X I , x2 1 0, two less than or equal to constraints
equations result. The coefficients of choice variables y of the dual are the transpose
of the first two columns of (4.3) to (4.6) giving
and
~ Y +4y2
I
+ loy4
(4.9)
In above constraint equations, the right-hand-side values are the first two coefficients of the primal objective.
I17
+ 4y4 = 200.
(4. IO)
Since the first two constraints in the primal are less than or equal to and since
the third constraint is greater than or equal to, we get
Y1,
y2, 5 0;
y3
20.
(4.1 1)
118
$A$2 XI
$B$2 x2
$C$2 x3
0
30
-57.8
232
0
0
82
100
200
1E+30
300
1E+30
232
1E+30
150
IConstraints
Cell
Name
$C$5 2x1+4x2 x3
Final
Value
$C$6 3 ~ 1 + 4 ~ 2 + ~3
6 ~ 3-226.8
$C$7 2x1-4x3 ~3
231.2
$C$8 1Ox2+5x3 x3
11
0
0
40
13
200
11
1E+30
31.2
39
239.8
1E+30
1E+30
Constraints
contains its dual formulation. The optimized objective in both solutions is the same,
namely -8560.
It is now instructive to examine the sensitivity reports of solutions, which are
reproduced in Figures 4.19, and 4.20.
119
For the primary problem, optimal choice variables are 0, 30, and -57.8. The
shadow prices corresponding to the four constraints are -75, 0, 0, and 40. In the
dual problem, values of shadow prices and choice variables are reversed: Final
values of choice variables are -75,0, 0, and 40, and shadow prices corresponding
to three constraints are 0, 30, and -57.8. The reader is asked to compare the
reduced costs and the allowable increase and decrease for variables for the two
solutions. What other conclusions can be drawn? Further by relating the solutions
to the duality theorems in Section 3.10.2, the reader will be in a position to discern
relations between the dual and the primal problems and will be able to draw some
interesting conclusions.
4.11
RESERVOIR MANAGEMENT
Linear programs are ideally suited to optimize problems associated with water flow
management, maximization of electricity generation in hydroelectric systems, and
a variety of other issues related to water resources management. Some problems,
particularly those involving the time of travel for water from one dam to another,
can be very complicated. The following modified example modeled after an exercise
in Rao (1996) is an illustration of a simple problem in water management.
120
Finally, from the ratio of outflows during wet and dry seasons, we obtain the
equality constraint as
The net flow into the imgation district during dry and wet seasons are xoutwet
0.7, and Xoutdry 0.1 respectively.
The objective is to minimize the cost given by
4.12
SUMMARY
The above examples and their setup on Excel sheets have exposed the reader to
ways of conceiving and structuring physical problems as optimization problems.
Clearly, one can set up ones spreadsheets in a manner different from those in these
spreadsheets. Nonetheless, the mathematical formulation of problems arising from
a physical phenomena is what the reader is expected to see and formulate in order
to obtain a solution.
121
4.13
EXERCISE PROBLEMS
1. The following problem illustrates an optimal mixing problem discussed in Section 4.1.
A generating station with several generators can obtain coal from three sources
with the following attributes:
Coal Type
A
B
C
% Sulfur
% Hg
MWNton
2.2
1.6
2.2
0.01
0.02
0.008
2
1.9
2.1
Cost, $/ton
42
45
40
The table indicates the percentage of pollutants, amount of energy (heat content)
per ton and the cost of coal.
(a) Find the optimal blending (procurement from) of supply from three sources
3. For the problem of Section 4.9.2, formulate its dual. Examine the sensitivity
sheet of the primal problem in Section 4.9.2 and compare it to the solution of
the dual problem. What is the meaning of the dual variables in this problem?
4. What is the increased cost of transporting the goods in the problem of Section 4.9.1 if
122
0
0
1\
150
300
Generation Area B
Import to Area D